The multiarm randomized comparative study design included 3 intervention groups that varied by human support mode: standard (S), comprising fully automated emails only; standard plus personalized SMS text messaging support (S+pSMS); standard plus videoconferencing support (S+VCS).

The study was advertised to members of a faith-based organization using a combination of methods (eg, email, social media, bulletins, and magazines), and interested individuals were directed to a website to apply. After randomization, eligible participants were emailed their group allocation, a participant information document, and instructions on how to complete registration onto the electronic learning management system (eLMS), signaling consent. Figure 1 demonstrates the flow of participants through the study.

The 10-week interdisciplinary intervention introduced participants to a range of evidence-based strategies for enhancing mental well-being, from the disciplines of lifestyle medicine and positive psychology. Participants logged onto the eLMS or mobile app to access the intervention, which included video content and a place to log daily and weekly experiential challenges (see screenshots in Multimedia Appendix 1). The website and mobile app were designed using a range of persuasive system design principles to increase engagement [15]. Each week, participants were encouraged to view one lesson, perform small daily challenges, and complete 1 larger weekly challenge. A detailed week-by-week summary of the content and experiential challenge activities can be found in Multimedia Appendix 2.

The study was structured to compare the value of increasingly greater levels of human support on adherence. Principles of the adherence-focused supportive accountability model [6] underpinned interactions between the support coach and participants. As the lowest form of support, one automated email was sent to the participants each week, consisting of the following: their first name as a salutation, a couple of sentences to encourage them to view the content, and a link to an introductory video by the presenter. A reminder email was sent if a participant had not logged experiential activity for 3 days, or viewed video content for 8 days. The second level of support involved SMS messages that were written and sent from a support coach who focused on ensuring content was process-focused; messages were sent regularly, and tone exuded positivity. Messages addressed the recipient by first name and were signed by the support coach. They were sent 3 times weekly for the first 3 weeks, and twice weekly for the remaining 7 weeks. In the highest level of support, the opportunity to develop key tenets of the supportive accountability model (eg, bond and legitimacy) was provided through weekly videoconferencing sessions (9 weekly time slots). Focus was given to facilitating a supportive, respectful environment where participants could share experiences safely. The facilitator sought to build trust through benevolent engagement with participants.

Analyses were conducted using SPSS, version 25 (IBM Corp). Preintervention-to-postintervention changes were calculated using paired t tests. Descriptive statistics, involving means and standard deviations, measured patterns of adherence. Analysis of variance (ANOVA) was used to compare adherence between groups. Fisher exact test was utilized to analyze relationships between categorical variables, and Cohen d measured effect size.

Avondale Human Research Ethics Committee granted ethics approval for the study (Approval No. 2018.09), and it was registered with the Australian and New Zealand Clinical Trial Registry (ANZCTR12619001009101). An email containing an “Information Statement” and a “Participant Consent Form” was sent to all prospective participants. The email explained that choosing to register on the eLMS would signify informed consent.

Potential participants (N=605) enrolled on the information website and agreed to participate in the study irrespective of randomization allocation. Subjects were randomized into 3 arms (S, n=201; S+pSMS, n=202; and S+VCS, n=202) (Figure 1). During a 1-week period, after being notified of group allocation, and before the intervention commenced, 24.3% (147/605) of participants dropped out. Early dropout attrition between groups was significantly different (P=.009). A total of 21.9% (44/201) withdrew from the S group, 19.3% (39/202) from the S+pSMS group, and 31.6% (64/202) from the S+VCS group. Table 1 shows that dropout was significantly different between groups when the S+VCS group was compared; differences were not significant between the S and S+pSMS groups.

A total of 75.7% (458/605) of randomized participants registered on the eLMS and completed the initial questionnaire (S, n=157; S+pSMS, n=163; S+VCS, n=138). Notably, there were no significant differences between the groups in age (P=.19), gender (P=.82), education (P=.16), or ethnicity (P=.34). Of the registered participants, 69.9% (320/458) completed the postintervention questionnaire (S, n=103; S+pSMS, n=114; S+VCS, n=103), resulting in 30.1% (138/458) being lost to follow-up; there was no difference between groups (P=.24).

Table 4 specifies the preferred mode of human support by group. A total of 88.7% (284/320) of participants indicated a valid human support preference. Almost half of participants within each group indicated a preference for combined support (ie, automated emails, personalized SMS, and VCS); however, the combined support alternative was not offered as part of the study. In comparison, the preference for solely S+VCS support was low across all the groups, with only between 7% to 8% choosing it from each group. Table 4 also indicates that 24% (22/91) in the S group, 27% (20/96) in the S+pSMS group, and 8% (8/97) in the S+VCS group received their preferred mode of support.

Secondary analysis was conducted to determine whether those who received their first preference for support mode scored better adherence or outcomes than those who did not receive their first choice. Participants (n=128) who chose the combined support as their first preference (an option not included in the comparative study) were excluded from the analysis, and the remaining participant data (n=156) were analyzed. Table 5 outlines the results between those who did not receive their first preference in human support mode and those who did. There was no significant difference between the 2 groups in either adherence or well-being measures.

Higher early dropout attrition occurred in the group which was allocated videoconferencing as a mode of human support. However, the mode of human support made no impact on attrition or adherence after the commencement of the intervention. Moreover, for participants who received their first preference in human support, compared to those who did not, no differences were observed in adherence. Preference for VCS support was low, yet almost half of the participants indicated they would prefer all forms of human support, though this was not an option in this study.

In this study, after the initial email notified participants of their group allocation, a disproportionate number of participants allocated to the S+VCS group withdrew from the study, with a significant between-group difference (P=.009). The early dropout attrition may have been influenced by dissatisfaction with the assigned support mode. Figure 1 shows that, in the S+VCS group, 57 participants either requested withdrawal or failed to register compared to 36 in the email group (S), and 28 in the S+pSMS group. Of the 24 participants who requested withdrawal from the S+VCS group, 10 stated that time was a factor, 8 provided no specific reason, 4 stated personal illness, 1 mentioned work commitments, and 1 participant stated they did not want to be involved in videoconferencing.

No differences in adherence between the groups were detected; and as documented in a previous report [39], we observed no differences between the groups in the well-being measures either. However, in observing that almost half (150/320, 46.9%) of the entire cohort were 100% adherent (ie, viewed all 10 videos), we conducted stratified analysis to compare well-being measures of those who were fully adherent (n=150) with those who were not (n=170). Significantly greater improvements were observed in the fully adherent group for life satisfaction (P=.011; d=0.15) and flourishing scores (P=.012; d=0.15), yet effect sizes were small.

While the primary aim of human support is usually to foster greater adherence [14,45,46], support did not influence adherence behavior for the participants in this study. While earlier research among clinical cohorts have indicated that supported interventions yielded better adherence and outcomes [29,47,48], numerous studies have also shown that human support made no difference [31,32,34,49-53]. Considering the extensive repertoire of previously identified factors affecting adherence, it is plausible that, in this nonclinical group, a wide range of variants may have influenced adherence at the individual level [14], reducing the potential impact of human support as a single factor. Therefore, exploring participants’ perceptions regarding influences on adherence is an important topic for further investigation.

The S+VCS was the least preferred support mode (7%), and low VCS attendance reflected the low preference for this type of support. Exploring reasons for the disinterest in videoconferencing as a mode of support may be a topic for further research, as videoconferencing has been used successfully in other group contexts [54-58] and has proven to be useful so long as technical support was provided [54]. Albeit, in a recent German study, 64% of patients were resistant to videoconferencing as a method of communication with health professionals. Notably, less than 1% reported previous experience with its use [59].

The S+VCS group allocation required an extra time commitment on behalf of participants compared to the other groups. Both email and SMS support required no effort by the participant—support was “pushed” to devices [7]. Conversely, videoconference support attempted to “pull” participants to an extra event, requiring effort and time to gain benefit from the support offered. Conceivably, the time and energy required to engage with that mode of support, loss of anonymity, technological barriers, unfamiliarity with videoconferencing software, or concerns about the group interaction may have been a barrier that facilitated significantly greater dropout. Additionally, further research should investigate if the preference for S+VCS would have been higher if support had been provided on an individual rather than a group basis.

For participants remaining in the S+VCS group, many demonstrated low engagement with the videoconferencing support. Consequently, by not engaging in the VCS support, they experienced a similar level of support to those in the S group (ie, automated emails only). This hindered the ability to draw meaningful between-group comparisons regarding the influence of human support offered through video conferencing on adherence.

Receiving their first preference in human support mode did not translate to participants reporting better adherence or outcomes in this study. We were unable to locate comparable studies for nonclinical groups. However, previous research among clinical cohorts has demonstrated that receiving support preferences may impact patient perceptions about the usefulness of an intervention [60] and improve adherence [28,61]. While a meta-analysis revealed that patients receiving their preferences demonstrated improved treatment satisfaction, adherence, and outcomes with moderate effect sizes [28], other research found that receiving the preferred option does not always impact outcomes [61,62]. A 2019 study of patients with anxiety and depression compared adherence and outcomes when patients chose their preferred support [63]. Interestingly, 78% chose the maximal support option, regular weekly support, and just 22% chose optional support (ie, support by request only). Yet, both groups achieved similar improvements in anxiety and depression scores and there were no differences in adherence. Contact between participants and therapist was much less for those who chose optional support, suggesting that similar results can be achieved with less time and cost investment.

Although combined human support (ie, access to automated emails, SMS, and videoconferencing) was not offered in the study, it is interesting that almost half of the participants in every group chose this as their preferred support option. While the reasons for this are unclear, it is hypothesized that it is a common trait of human nature to want access to everything possible, even though the available options are not necessarily utilized or needed. Further research exploring participant perceptions regarding human support modes and preferences is warranted.

The intervention and its implementation were supported by established theory. While the intervention itself was underpinned by the theory of planned behavior [64], the design components of the intervention for the web- and mobile app–based platforms were informed by the persuasive design model [15] and human support elements reflected principles of the supportive accountability model [6]. The use of a sole facilitator as a support person provided consistency in the participant’s experience with regards to technical assistance, messaging, and videoconference facilitation. The study attracted a large cohort with a broad range of ages (18-81 years); and adherence data was easily collected through the eLMS and mobile app, avoiding the possibility of human error. Despite the addition of human support elements, the administration of the intervention using the eLMS and the mobile app provided acceptable scalability, portability, and accessibility, demonstrating the potential for broad-based mental health promotion.

Several limitations should be noted. The study attracted mainly White, well-educated women. While this is often seen in digital health interventions [65,66], it limits the ability to generalize more widely. Data gathering relied on self-reporting, which may be subject to bias, and self-selection into the study may have resulted in a cohort skewed by factors such as technological ability and motivation to achieve better mental health. Bias may have been introduced by the disproportionate number of participants who withdrew from the S+VCS group after being notified of their group allocation, reducing the validity of the randomization process. In asking participants to rank human support preferences, analyses were limited by including a preference option in the questionnaire that was not included in the study (ie, combined support involving email, SMS, and VCS). Almost half of the participants ranked combined support as their first preference and were consequently excluded from further investigation, reducing the power of the analyses. Additionally, some factors that would have been useful for group comparison were not measured. For example, asking participants to indicate their level of engagement with SMS and emails, and collecting data regarding preference for use of the mobile app compared to the web-based experience, would have provided an indication of engagement with the human support offered.

The findings of this study indicate that a web- and mobile app–based MHPI for a nonclinical cohort can be designed and implemented to maximize accessibility, scalability, and adherence without the additional cost of human support. While early dropout attrition may have been influenced by displeasure with allocated support, adherence to a 10-week MHPI for a healthy cohort was not impacted by differing modes of human support. Engagement with videoconference support was suboptimal, hindering the ability to draw meaningful between-group comparisons. However, SMS support demonstrated no added value compared to automated email support. Adherence was not impacted by participants receiving their first preference for support. Future research should explore participant perspectives on adherence behaviors.